Quantcast

An Image Classification Algorithm and its Parallel Implementation Based on ANL-RBM

Research paper by Song, Haifeng, Chen, Guangsheng, Yang, Weiwei

Indexed on: 19 Jul '18Published on: 01 Jul '18Published in: Journal of Information Technology Research



Abstract

This article describes how when using Restricted Boltzmann Machine (RBM) algorithm to design the image classification network. The node number in each hidden layer, and the layer number of the entire network are designed by experiments, it increases the complexity for the RBM design. In order to solve the problem, this article proposes an image classification algorithm based on ANL-RBM (Adaptive Nodes and Layers Restricted Boltzmann Machine). The algorithm can automatically calculate the node number in each hidden layer and the layer number of the entire network. It can reduce the complexity for the RBM design. In the meantime, this article has designed the parallel model of the algorithm in the Hadoop platform. The experimental results showed that the image classification algorithm based on an ANL-RBM has a higher execution efficiency, better speedup, better scalability and it is suitable for massive amounts of image data processing.